FreeNAS Box v2: Construction Complete

After spending an afternoon + evening at a Tux-Lab work session, I have my FreeNAS Box v2! It’s always fun to see my idea turned into reality.

FreeNASv2 Complete

The first thing I appreciated was the fact that the components are clearly visible through the acrylic panels. And even better, the messy tangle of wires are hidden behind them. This reversal from v1 is the best aesthetic change.

The other major design requirement – that both cooling fans be visibly spinning – is also present but it doesn’t have as much of an immediate aesthetic effect.

After I’ve kept it on and running overnight, I checked the temperature around the box the following morning. I think it would be neat to check thermal performance with a FLIR thermal imaging camera but lacking such toys I went with the low-tech way of putting my hands at various places around the box to feel the temperature.

The front chamber – where the CPU and motherboard reside – has a slight temperature gradient from top to bottom but overall it was relatively cool to the touch. This was expected as the CPU basically sat idle all night. It also means I won’t need to cut a hole in the front door for a direct air intake.

The rear chamber, with the power supply and both hard drives, is where most of the heat is generated. The two drives were warm to the touch signifying that they’ve been spinning all night and getting some amount of cooling to keep them from getting hot.

The power supply fan was running and the power supply case was cool to the touch. The power meter read 30W for the FreeNAS box in this steady-state idle state. This is a very light load for a 600W-rated power supply, reflected in its cool running temperature.

Vacuum Table – Spoilboard and Gasket

Now that we have a baseline on the vacuum table performance, time to start performing modifications to see what happens.

The easiest thing to reduce air resistance is to remove layers – we don’t strictly need the spoilboard in this setup, so it is removed. We then added some rubber gaskets to improve the seal between the plenum and the fixture, which should reduce air leaking past that particular junction.

8 Plenum Gasket Fixture small
Spoil board removed, gasket added – 25 inches

These modifications did not drop the vacuum of an empty fixture – in fact vacuum was boosted by 2 inches to 25. This implied having the spoilboard in the setup was letting a lot of air slip around the fixture. Removing it was a good call.

9 Plenum Gasket Fixture Blanks small
Adding work pieces increased to 26 inches.

When the work pieces are in place, the vacuum went up another inch to 26 inches. Less air is leaking past the work pieces, and they are now held by about one inch of mercury (roughly half a pound per square inch.)

We haven’t put any effort into improving the sealing between the work pieces and the fixture. How much gain can we realistically expect from the effort? In order to get a rough estimate of how much more we might gain, we draped a plastic sheet over the fixture.

10 Plenum Gasket Fixture Sheet small
Replacing work pieces with a plastic sheet boosted to 27.5 inches.

Looks like we have about 1.5 inches of mercury we can gain from better workpiece-to-fixture sealing.

This is a promising start, as this tells us we’re in reach of a decently high level of vacuum for work-holding. We now need to put some effort into the other side – improve the path for the vacuum to reach the work pieces and hold them in place.

More improvements to come!

Vacuum Table – Baseline Measurements

The CNC router at Tux-Lab has been under-utilized partly due to its under-performing vacuum table. It has a poor track record on an existing project, and we want to understand why (and hopefully fix the problem) before doing more projects on the CNC router.

To narrow down the cause, we will record the pump’s vacuum gauge reading at various configurations. We use a phone to take a picture identifying the vacuum configuration. We then hold that picture up next to the gauge and take a picture of the phone and the gauge together.

Establish the bounds

First, we get the upper bound: once the pump is up and running, close all the zone valves. The reading – nearly 30 inches of mercury – confirms the pump itself and the majority of the vacuum piping is in good working order.

2 Zone Valves small
Maximum vacuum – nearly 30 inches, good!

The lower bound is obtained by opening all zone valves and place nothing on the spoilboard. When in working configuration, the vacuum will never be weaker than this 7.5 inch reading.

6 Spoilboard small
Minimum vacuum – roughly 7.5 inches.

Most of the tests confirmed that the vacuum setup itself appears to be in good working order. We only started seeing problematic numbers once we started involving the spoilboard and the project fixture. Good news since these are the easiest pieces to fix.

4 Spoilboard Fixture small
Fixture without work pieces – just under 23 inches.

Past runs of the existing project has been done with the fixture mounted on the spoilboard. The vacuum reading of this configuration is surprisingly high at a hair under 23 inches. Indicating a lot of air resistance despite being carved from low density fiberboard.

We then added the work piece blanks on top of the fixture and measured again.

5 Spoilboard Fixture Blanks small
Fixture with work pieces – just over 23 inches.

The vacuum barely changed, to just a hair over 23 inches. This is a problem: it tells us the air can easily find a path around the work pieces so very little of atmospheric pressure is applied to hold pieces in place.

Now the objective is to modify the setup to (1) reduce the vacuum reading of an empty fixture and also (2) increase the vacuum reading of the populated fixture.

Increasing difference between these two readings should increase the holding power.

Mini-ITX Server Box

Mini-ITX Server CADTux-Lab had components on hand for a completely fan-less bare-bones Mini-ITX system. A small board with a passively-cooled CPU, a small 12V DC to ATX DC power supply that didn’t need a fan, and a solid state drive for storage. All it needed was a simple basic box to keep everything in – which made it an ideal learning project as a follow-on to FreeNAS box V1. (Well, it can be argued that this simple box should have come first… but that wasn’t the order things ended up being.)

This time there was no design challenge with hard drive placement or power supply fan clearance. Just a simple box with two sets of holes so convection will pull cold air from the bottom and let hot air out the top. The back plate had opening for the standard ITX motherboard port plate, plus two holes: One for the 12V DC power input, and the other for a momentary-on power switch.Mini-ITX Server

The result was an upgrade from its previous placement, which was the bare circuit board sitting on top of a cardboard box. Now it has some minimal protection against accidents like an errant dropped paperclip shorting things out.

This machine is now set up with the Xen hypervisor and ready to run the server-side code of whatever future projects arise at Tux-Lab, as long as that code can run in a Xen virtual machine.

FreeNAS Box V1 Prototype

FreeNASv1With the concept designed, it’s time to head over to Tux-Lab to build it!

To be honest, it was not fully designed for use as a computer case. Since this was my first effort designing for acrylic construction I expected to run into some amateur mistakes very quickly. As a result I had left some known design issues open to be resolved in future prototypes. One example: I had not designed any kind of door or hinge. The prototype panels are mostly glued together, except for the front panel which is held in place by tape.

Aesthetically, I am pleased with how the clear acrylic looked though I am not pleased at how much of a rat’s nest the power supply cables became. Taming wires is a perpetual challenge. I now understand why Apple enclosed all the ugly guts of the G4 cube in shiny aluminum shell inside the clear acrylic shell.

Other than the messy computer wires, the clear acrylic does hint at the illusion of a computer floating in mid air. I’m pretty happy with that, but it’s not enough to offset the tangle of wires. Next prototype will either have good cable management (takes effort) or have some dark colored acrylic to hide the interior (much easier.)

The cooling functionality worked as designed: the intake air is drawn from the bottom, past the two hard drives keeping them cool, and out the top.

Similarly, the designed goals of tilted-PSU (power supply unit) space optimization was successful, as did the gentler turn demanded of the wires. However, there was an unforeseen deal-breaker of an issue.


FreeNASv1_FlawOn the back side of the tilted-PSU, we see that the tilt has pressed the bottom of the PSU up against the wire bundle at the top of the motherboard. The tight quarters mean individual wires of the bundle tried to relieve the crowding by moving into the space for the PSU fan preventing it from turning. Since the PSU fan is the primary air-mover for this enclosure, a stopped fan is obviously not acceptable.

Other notes

Space utilization efficiency has room for improvement. Some of this was caused by the desire to emulate the Apple G4 cube and have a square footprint. (20 cm squared!) The squareness was completely unnecessary and future iteration will likely have a rectangular footprint for space efficiency.

There was uncertainty about how well 3mm acrylic can hold the weight of the power supply unit. It proved to be surprisingly capable once the two top sheets reinforced each other at right angle.

Amateur Hour: A laser cutter only cuts vertically. There was no way to cut a 30 or 60 degree edge for the tilted PSU section! For this learning exercise, the cornered edges are simply left open and unattached.

The angled PSU was a novel idea to solve specific problems, but it caused new ones and also unsuitable for laser cut acrylic construction. That was a fun experiment, but we’ll have to leave the angled PSU concept behind for the next prototype.

Luggable Frame Experiment #2

Catleap2The second iteration of the luggable frame experiment addressed the failings of the first version by relying less on acrylic and more on aluminum. The first iteration was a good experiment to see if acrylic was strong enough for the work. Once V1 conclusively proved the weaknesses, it’s time to fall back to the known quantity.

The following changes were made for version 2:

Extrusion upgrade: In the interest of greater rigidity, the extrusions themselves were upgraded from Misumi HFS3 (15mm x 15mm cross section) to HFS5 (20mm x 20mm). The smaller extrusions seemed to be doing the job but they did exhibit some flex. And we had HFS5 conveniently on hand so let’s use it!

Connection upgrade: In V1 the extrusion T-joint at the base of the frame was held together by the side pieces of acrylic. Though it seemed to work, V2 went with a stronger solution by using metal connectors for the joint. (Misumi HBLFSNF5).

Handle upgrade: The V1 handles were part of the acrylic assembly. With the reduction in acrylic usage, there wasn’t enough left to carry the load of the whole frame. So the handle became another aluminum extrusion.

Catleap2-RearPC tray upgrade: This was the first acrylic thing that failed in V1. The PC is now held in place by aluminum structure instead of an acrylic cutout which makes it quite secure. Three of the extrusion right-angle connectors were re-purposed as “claws” to keep the PC case in place.

Catleap2-SideVESA mount upgrade: The worrisome flex in the Catleap monitor enclosure was traced down to the metal threads inside the Catleap enclosure that were longer than the thickness of the enclosure plastic. This meant when the mounting screws fully engaged, there was still a bit of space between the VESA mount plate and the monitor’s rear surface, allowing movement. A spacer plate was added to fill that gap. Now the VESA mounting plate on the frame is fully pressed against the monitor’s rear surface, greatly reducing the flex.

All this additional structure added up to a very secure frame for carrying around the Yamakasi Catleap monitor with the HP Z220 computer. Unfortunately it also added weight which was a concern even before the frame came into the picture. The heft means this is probably the end of the line for the Catleap + Z220 experiments. Frame V2 will serve as a perfectly good workstation albeit not a very portable one.

The idea of building a Luggable PC around a commercially available monitor will continue, with the focus shifting to using smaller and lighter components.

Luggable Frame Experiment #1

Catleap1The dimensions for my Luggable PC project were determined by the components within. The width and height, specifically, were dictated by the LCD screen module. Even though I made the CAD files public for anybody to build their own Luggable PC, in practical terms only people with the exact same LCD module would be able to use the files without modification.

A friend who saw the Luggable PC was interested in generalizing the concept and create a frame for lugging a (not disassembled) screen alongside its (also not disassembled) PC. Relative to my project, it would be easier to build and less specialized to the components within, with a trade-off in larger size and heavier weight.

I thought it was a great idea to explore and joined in the experiment. We each came up with a design, and we built both of them at Tux-Lab to see how the ideas translated into reality.

This blog post is a brief summary of my first experiment.

The Components

The monitor is an Yamakasi Catleap monitor, built around a 27″ IPS panel with 2560×1440 resolution. The specific dimensions don’t really matter, as it will be mounted via the standard 75mm VESA pattern on the back. Any large monitor with 75mm VESA pattern would fit as-is, and only minor modifications would be necessary to accommodate monitors with a different mounting pattern.

The PC is a HP Z220, small form factor PC from the HP business line available with a range of components to trade off processing power against price. For the purposes of this experiment, the important details are its height of 331mm and depth of 100mm. Thought not a standardized dimension, many small form factor PCs are roughly the same size.

The Construction

The core of the frame are built from 15mm aluminum extrusions (Misumi HFS3) for strength and the remainder of the frame are made from 6mm laser-cut acrylic fastened to the extrusions via M3 nuts and bolts.

Making the panels from laser-cut acrylic has the advantage of simpler modifications. Many of the critical dimensions in my Luggable PC 3D CAD file has the problem that, when changed, they trigger cascading changes that need to be reconciled. When designing for the 2D tool path of laser laser cutting, it is easier to keep modifications in mind so that a change in one sheet does not cascade to other sheets.

Example #1: The frame has a 331mm x 100mm hole to fit the Z200 case. This can be adjusted to fit any other SFF frame without cascading changes to other components.

Example #2: The monitor mount pattern can be changed, and the mount position can be moved up or down to adjust elevation of the monitor.

The Result

CompleteI had never designed for laser cutting before and was happy for the chance to do something on the Tux-Lab laser cutter. I knew that, having little experience with the material, my first few designs will have some amateurish flaws. So this frame #1 was fairly minimalist just to see what happens.

I didn’t have a good grasp how many fasteners I would need to hold everything together. I laser-cut roughly double the number of fastener positions than what I think I would need, as it is easier to have more options rather than less. For the assembly I only installed fasteners in every other hole.

The screen mount was surprisingly successful. We questioned whether 6mm acrylic would be suitable for holding up the Catleap monitor by its 75mm VESA mounts. When we found some worrisome flex, the suspicion went immediately to the 6mm acrylic but it turned out the Catleap monitor enclosure was the source of the flex.

When attempting to install the PC, we found that the case itself would fit just fine but the rubber feet attached to the side of the case did not. I added cutouts in the CAD file but it seemed wasteful to cut entirely new pieces of acrylic just for the little feet cutouts. For purposes of experimentation, a Dremel tool was used to cut gaps to clear the rubber feet.

After the frame was assembled with the screen and the PC, we started plugging in all the cables and wires and I realized I had forgotten to account for the cables. There’s no good place to coil up the excess so they kind of dangle and stand ready to catch on something inconvenient.

The entire assembly was built in a tiny fraction of the time of my Luggable PC and included a much larger monitor with a much higher resolution. The trade off was almost doubling of the weight. The handle, part of the acrylic assembly, appeared to be sufficient to manage the weight.

I carried it across Tux-Lab and quickly encountered the first failure.

The Failure

Lesson of the Day: Sharp internal corners are bad.

My amateur mistake was cutting a sharply cornered rectangle to hold the PC. The sharp corners concentrated the physical load of the PC into a small point in the 6mm acrylic, which protested the poor design by breaking apart.

The next experiment will incorporate this lesson.

Build, fail, learn, iterate, repeat.